LIFELONG EDUCATION January 14Th 2017

Home , The Economist, The Economist (Lost)

SPECIAL REPORT LIFELONG EDUCATION January 14th 2017

Learning and earning

20170114_Education.indd 1 15/12/2016 18:55 SPECIAL REPORT LIFELONG EDUCATION

Learning and earning

Technological change demands stronger and more continuous connections between education and employment, says Andrew Palmer. The faint outlines of such a system are now emerging THE RECEPTION AREA contains a segment of a decommissioned Un- CONTENTS derground train carriage, where visitors wait to be collected. The surfaces are wood and glass. In each room the talk is of code, web development 3 The role of employers and data science. At first sight the London office of General Assembly Cognition switch looks like that of any other tech startup. But there is one big difference: whereas most firms use technology to sell their products online, General 4 The ageing brain Old dog, new tricks Assembly uses the physical world to teach technology. Its office is also a campus. The rooms are full of students learning and practising code, 5 Upstarts and incumbents many of whom have quit their jobs to come here. Full-time participants The return of the MOOC ACKNOWLEDGMENTS have paid between £8,000 and £10,000 ($9,900-12,400) to learn the lin- gua franca ofthe digital economy in a programme lasting10-12 weeks. 8 Career planning Apart from those mentioned in Pathway dependency the text, thanks for their help in General Assembly, with campuses in 20 cities from Seattle to Syd- preparing this report are due to: ney, has an alumni body of around 35,000 graduates. Most of those who 9 Low-skilled workers David Autor, Bill Blase, Anna enroll for full-time courses expect them to lead to new careers. The com- The elephant in the truck Breman, Brian Callaghan, Tom Davenport, Shernaz Daver, Rasmus pany’s curriculum is based on conversations with employers about the Enemark, Karen Evans, Charles skills they are critically short of. It holds “meet and hire” events where Fadel, Jose Ferreira, Jonathan firms can see the coding work done by its students. Career advisers help Finkelstein, Alison Fuller, Art students with their presentation and interview techniques. General As- Graesser, Guy Halfteck, Kathleen Hogan, Peter Horrocks, Denis sembly measures its success by how many ofits graduates get a paid, per- Hurley, Jerry Kaplan, Ewart Keep, manent, full-time job in their desired field. Ofits 2014-15 crop, three-quar- Nathan Kuncel, Rose Luckin, ters used the firm’s career-advisory services, and 99% ofthose were hired Christian Lyhne Ibsen, Robert within 180 days ofbeginning their job hunt. Lyngman, Michael Moe, Mona Mourshed, Liz Rees, Laurent The company’s founder, Jake Schwartz, was inspired to start the

Reich, Bart Rienties, Sanjay companybytwo personal experiences: a spell ofdriftingafter he realised

¢ ¡ Sarma, Stefano Scarpetta, David that his degree fr ale conferred no practical skills, and a two-year Schejbal, Nicole Smith, Mitchell MBA that he felt had cost too much time and money: “I wanted to change Stevens, Andrew Tolmie, Anders Vind, David Willetts, Dan Willing- the return-on-investment equation in education by bringing down the A list of sources is at ham and Jonathan Zimmerman. costs and providing the skills that employers were desperate for.” 1 Economist.com/specialreports

The Economist January 14th 2017 1 SPECIAL REPORT LIFELONG EDUCATION

This analysis buttresses the view that technology is already playing havoc with employment. Skilled and unskilled workers alike are in trouble. Those with a better education are still more likely to find work, but there is now a fair chance that it will be unenjoyable. Those who never made it to college face being squeezed out ofthe workforce altogether. This is the argument of the techno-pessimists, exemplified by the projections of Carl-Be- nedikt Frey and Michael Osborne, of Oxford University, who in 2013 famously calculated that 47% ofexisting jobs in America are susceptible to automation. There is another, less apocalyptic possibility. James Bessen, an economist at Boston University, has worked out the effects of automation on specific professions and finds that since 1980 employment has been growing faster in occupations that use computers than in those that do not. That is because automation tends to affect tasks within an occupation rather than wiping out jobs in theirentirety. Partial automation can actually increase demand by reducing costs: despite the introduction of the barcode 2 In rich countries the linkbetween learning and earning has scanner in supermarkets and the ATM in banks, for example, the tended to follow a simple rule: get as much formal education as number ofcashiers and banktellers has grown. you can early in life, and reap corresponding rewards for the rest But even though technology may not destroy jobs in aggre- ofyourcareer. The literature suggeststhateach additional yearof gate, it does force change upon many people. Between 1996 and schooling is associated with an 8-13% rise in hourly earnings. In 2015 the share of the American workforce employed in routine the period since the financial crisis, the costs of leaving school office jobs declined from 25.5% to 21%, eliminating 7m jobs. Ac- early have become even clearer. In America, the unemployment cording to research by Pascual Restrepo of the Massachusetts In- rate steadily drops as you go up the educational ladder. stitute of Technology (MIT), the 2007-08 financial crisis made Many believe that technological change only strengthens things worse: between 2007 and 2015 job openings for unskilled the case for more formal education. Jobs made up of routine routine worksuffered a 55% decline relative to other jobs. tasks that are easy to automate or offshore have been in decline. In many occupations it has become essential to acquire The usual flipside of that observation is that the number of jobs new skills as established ones become obsolete. Burning Glass requiring greater cognitive skill has been growing. The labour Technologies, a Boston-based startup that analyses labour mar- market is forking, and those with college degrees will naturally kets by scraping data from online job advertisements, finds that shift into the lane that leads to higher-paying jobs. the biggest demand is for new combinations of skills—what its The reality seems to be more complex. The returns to edu- boss, Matt Sigelman, calls “hybrid jobs”. Coding skills, for exam- cation, even for the high-skilled, have become less clear-cut. Be- ple, are now being required well beyond the technology sector. tween 1982 and 2001 the average wages earned by American In America, 49% of postings in the quartile of occupations with workers with a bachelor’s degree rose by 31%, whereas those of the highest pay are for jobs that frequently ask for coding skills high-school graduates did not budge, according to the New York (see chart). The composition ofnew jobs is also changing rapidly. Federal Reserve. Butin the following12 yearsthe wagesofcollege Over the past five years, demand for data analysts has grown by graduates fell by more than those of their less educated peers. 372%; within that segment, demand for data-visualisation skills Meanwhile, tuition costs at universities have been rising. has shot up by 2,574%. Acollege degree at the start ofa workingcareerdoes not an-

A question of degree, and then some swer the need for the continuous acquisition of new skills, espe- ¤

The decision to go to college still makes sense for most, but ciallyascareerspansare lengthening £ocational training isgood the idea of a mechanistic relationship between education and at giving people job-speciﬁc skills, but those, too, will need to be wages has taken a knock. A recent survey conducted by the Pew updated over and over again during a career lasting decades. Research Centre showed that a mere 16% ofAmericans thinkthat “Germany is often lauded for its apprenticeships, but the econ- a four-year degree course prepares students very well for a high- omy has failed to adapt to the knowledge economy,” says An-

paying job in the modern economy. Some of this may be a cycli- dreas Schleicher, head of the education directorate of the OECD,

¦¨©

cal effect ofthe financial crisis and its economic aftermath. Some a club of mostly rich coun £ocational training has a role, of it may be simply a matter of supply: as more people hold col- buttrainingsomeone earlyto do one thingall theirlives isnotthe lege degrees, the associated premium goes down. But technol- answer to lifelong learning.” 1 ogy also seems to be complicating the picture. A paper published in 2013 by a trio of Canadian economists, Paul Beaudry, David Green and Benjamin Sand, questions optimistic assumptions about demand for non-routine work. In the two decades prior to 2000, demand for cognitive skills soared as the basic infrastructure of the IT age (computers, serv- ers, base stations and fibre-optic cables) was being built; now that the technology is largely in place, this demand has waned, say the authors. They show that since 2000 the share ofemploy- ment accounted forby high-skilled jobs in America has been fall- ing. As a result, college-educated workers are taking on jobs that are cognitively less demanding (see chart), displacing less educated workers.

2 The Economist January 14th 2017 2 Such specific expertise is meant to be acquired on the job, but employers seem to have become less willing to invest in trainingtheirworkforces. In its2015 EconomicReportofthe Presi- dent, America’s Council of Economic Advisers found that the share of the country’s workers receiving either paid-for or on- the-job training had fallen steadily between 1996 and 2008. In Britain the average amount of training received by workers al- most halved between 1997 and 2009, to just 0.69 hours a week. Perhaps employers themselves are not sure what kind of expertise they need. But it could also be that training budgets are particularly vulnerable to cuts when the pressure is on. Changes in labour-market patterns may play a part too: companies now have a broader range of options for getting the job done, from automation and offshoring to using self-employed workers and crowdsourcing. “Organisations have moved from creating talent to consuming work,” says Jonas Prising, the boss of Manpower, an employment consultancy. Add all of this up, and it becomes clear that times have got tougher for workers of all kinds. A college degree is still a prere- quisite for many jobs, but employers often do not trust it enough to hire workers just on the strength of that, without experience. In many occupations workers on company payrolls face the prospectthattheirexistingskillswill become obsolete, yetitis often not obvious how they can gain new ones. “It is now reason- able to ask a marketing professional to be able to develop algorithms,” says Mr Sigelman, “but a linear career in marketing doesn’t offeran opportunity to acquire those skills.” And a growing number ofpeople are self-employed. In America the share of The role of employers temporary workers, contractors and freelancers in the workforce rose from 10.1% in 2005 to 15.8% in 2015. Cognition switch Reboot camp The answer seems obvious. To remain competitive, and to give low- and high-skilled workers alike the best chance of success, economies need to offer training and career-focused education throughout people’s working lives. This special report will Companies are embracing learning as a core skill chart some of the efforts being made to connect education and employment in new ways, both by smoothing entry into the la- A STRANGE-LOOKING SMALL room full of vintage furni- bourforce and byenablingpeople to learn newskillsthroughout ture—an armchair, a chest of drawers, a table—was being their careers. Many of these initiatives are still embryonic, but built in the middle ofInfosys’s Palo Alto offices when your corre- they offer a glimpse into the future and a guide to the problems spondent visited in November. Tweed jackets hung from a raised by lifelong reskilling. clothes rack; a piano was due to be delivered shortly. The struc- Quite a lot is already happeningon the ground. General As- ture was rough and unfinished. And that, accordingto Sanjay Ra- sembly, forexample, is just one ofa numberofcoding-bootcamp jagopalan, was largely the point. providers. Massive open online courses (MOOCs) offered by MrRajagopalan ishead ofresearch and design atthe Indian companies such as Coursera and Udacity, feted at the start ofthis business-services firm. He is a disciple of “design thinking”, a decade and then dismissed as hype within a couple of years, problem-solving methodology rooted in observation ofsuccess- have embraced new employment-focused business models. ful innovators. His goal is an ambitious one: to turn a firm that LinkedIn, a professional-networkingsite, bought an online train- built a global offshoring business by following client specifica- ingbusiness, Lynda, in 2015 and isnowofferingcoursesthrough a tions into one that can set the terms ofits projects for itself. service called LinkedIn Learning. Pluralsight has a library of on- Design thinking emphasises action over planning and en- demand training videos and a valuation in unicorn territory. courages its followers to lookat problems through the eyes ofthe Amazon’s cloud-computing division also has an education arm. people affected. Around 100,000 Infosys employees have gone Universities are embracing online and modular learning through a series ofworkshops on it. The first such workshop sets more vigorously. Places like Singapore are investing heavily in the participants a task: forexample, to improve the experience of providing their citizens with learning credits that they can draw digital photography. That involves moving from the idea ofmak- on throughout their working lives. Individuals, too, increasingly ing a better camera to considering why people value photo- seem to accept the need for continuous rebooting. According to graphsin the firstplace, asa wayofcapturingmemories. Asideas the Pew survey, 54% of all working Americans think it will be es- flow, people taking part in the workshops immediately start pro- sential to develop new skills throughout their working lives; ducing prototypes with simple materials like cardboard and pa- among adults under 30 the number goes up to 61%. Another sur- per. “The tendency is to plan at length before building,” says Mr vey, conducted byManpowerin 2016, found that93% ofmillenni- Rajagopalan. “Our approach is to build, build, build, test and als were willing to spend their own money on further training. then plan.” Meanwhile, employers are putting increasing emphasis on That baffling structure in Palo Alto was another teaching learning as a skill in its own right. 7 tool. MrRajagopalan had charged a small team with reimagining 1

The Economist January 14th 2017 3 SPECIAL REPORT LIFELONG EDUCATION

2 the digital retail experience. Instead of coming up with yet another e-commerce Old dogs, new tricks site, they were experimenting with technologies to liven up a physical space. (If a weary shopper sat in the chair, say, a pot As people age, the brain changes in both good ways and bad of tea on an adjacent table would auto- matically brew up.) The construction of IF YOU ARE over 20, look away now. Your hypothesis, but a range of tentative explana- the shop prototype in Infosys’s oﬃces cognitive performance is probably already on tions has been put forward. One points the was being documented so that employ- the wane. The speed with which people can ﬁnger at myelin, a white, fatty substance that ees could see design thinking in action. process information declines at a steady rate coats axons, the tendrils that carry signals Infosys is grappling with a vital from as early as their 20s. from one neuron to another. Steady reduc- question: what do people need to be A common test of processing speed is tions in myelin as people age may be slowing good at to succeed in theirwork? Whatev- the “digit symbol substitution test”, in which down these connections. Another possibility, er the job, the answer is always going to a range of symbols are paired with a set of says Timothy Salthouse, director of the

involve some technical and speciﬁc skills, numbers in a code. Participants are shown the Cognitive Ageing Laboratory at the University

based on knowledge and experience of a code, given a row of symbols and then asked o irginia, is depletion of a chemical called particular industry. But with design to write down the corresponding number in dopamine, receptor sites for which decline in thinking, Infosys is focusing on “founda- the box below within a set period. There is number with advancing age. tional skills” like creativity, problem-solv- nothing cognitively challenging about the Fortunately, there is some good news to ing and empathy. When machines can task; levels of education make no difference go with the bad. Psychologists distinguish put humans to shame in performing the to performance. But age does. Speed consis- between “fluid intelligence”, which is the routine job-specific tasks that Infosys tently declines as people get older. ability to solve new problems, and “crystal- once took offshore, it makes sense to Why this should be is still a matter of lised intelligence”, which roughly equates to think about the skills that computers find an individual’s stock of accumulated know- harder to learn. ledge. These reserves of knowledge continue David Deming of Harvard Universi- to increase with age: people’s performance ty has shown that the labour market is al- on vocabulary and general-knowledge tests ready rewarding people in occupations keeps improving into their 70s. And experi- that require social skills. Since 1980 ence can often compensate for cognitive growth in employment and pay has been decline. In an old but instructive study of fastest in professions across the income typists ranging in age from 19 to 72, older scale that put a high premium on social workers typed just as fast as younger ones, skills (see chart, next page). even though their tapping speed was slower. Social skills are important for a They achieved this by looking further ahead wide range of jobs, not just for health- in the text, which allowed them to keep going care workers, therapists and others who more smoothly. are close to their customers. Mr Deming What does all this mean for a lifetime of thinks their main value lies in the rela- continuous learning? It is encouraging so tionship between colleagues: people long as people are learning new tricks in who can divide up tasks quickly and ef- familiar fields. “If learning can be assimilated fectively between them form more pro- into an existing knowledge base, advantage ductive teams. If work in future will in- tilts to the old,” says Mr Salthouse. But creasingly be done by contractors and moving older workers into an entirely new freelancers, thatcapacityforco-operation area of knowledge is less likely to go well. will become even more important. Even geeks have to learn these skills. Ryan Ros- lansky, who oversees LinkedIn’s push into online education, notes that many software engineers are clude Pratt & Whitney, a maker of aircraft engines, and Otis, a lift taking management and communications courses on the site in manufacturer. Since 1996 UTC has been running a programme order to round themselves out. under which its employees can take part-time degrees and have tuition fees of up to $12,000 a year paid for them, no strings at- Building a better learner tached. Employers often balkat training staffbecause they might Another skill that increasingly matters in finding and keep- leave for rivals, taking their expensively gained skills with them. ing a job is the ability to keep learning. When technology is But Gail Jackson, the firm’s vice-president of human resources, changing in unpredictable ways, and jobs are hybridising, hu- takes a different view. “We want people who are intellectually mans need to be able to pickup new skills. At Infosys, Mr Rajago- curious,” she says. “It is better to train and have them leave than palan emphasises “learning velocity”—the process ofgoing from not to train and have them stay.” a question to a good idea in a matter of days or weeks. Eric Such attitudes are becoming more common. When Satya Schmidt, now executive chairman of Alphabet, a tech holding Nadella took over as boss of Microsoft in 2014, he drew on the company in which Google is the biggest component, has talked workofCarol Dweck, a psychologyprofessoratStanford Univer- ofGoogle’s recruitment focus on “learning animals”. MarkZuck- sity, to push the firm’s culture in a new direction. Ms Dweck di- erberg, one of Facebook’s founders, sets himself new personal vides students into two camps: those who thinkthat ability is in- learning goals each year. nate and fixed (dampening motivation to learn) and those who An emphasis on learning has long been a hallmark of Un- believe that abilities can be improved through learning. This ited Technologies (UTC), a conglomerate whose businesses in- “growth mindset” is what the firm is trying to encourage. It has 1

4 The Economist January 14th 2017 SPECIAL REPORT LIFELONG EDUCATION

customers in the order in which they arrive in the game, for example, might serve as an indicator of integrity. Intellectual curiosity is one ofthe traits that Knacktests for. The second question iswhetheritispossible to train people to learn. Imaging techniques are helping unlock what goes on in the mind of someone who is curious. In a study published in 2014 in Neuron, a neuroscience journal, participants were first asked to rate their curiosity to learn the answers to various questions. Laterthey were shown answers to those questions, as well as a picture of a stranger’s face; finally, they were tested on their recall of the answers and given a face-recognition test. Greater curiosity led to better retention on both tests; brain scans showed increased activity in the mesolimbic dopamine system, a reward pathway, and in the hippocampus, a region that matters for forming new memories. It is too early to know whether traits such as curiosity can be taught. But it is becoming easier to turn individuals into more effective learners by making them more aware of their own thought processes. Hypotheses about what works in education 2 amended its performance-review criteria to include an appraisal and learning have become easier to test because ofthe rise ofon- of how employees have learned from others and then applied line learning. MIT has launched an initiative to conduct interdis- that knowledge. It has also set up an internal portal that inte- ciplinary research into the mechanics of learning and to apply grates Lynda, the training provider bought by LinkedIn (which the conclusions to its own teaching, both online and offline. It Microsoft itselfis now buying). uses its own online platforms, including a MOOC co-founded AT&T, a telecoms and media firm with around 300,000 with Harvard University called edX, to test ideas. When MOOC employees, faces two big workforce problems: rapidly changing participants were required to write down their plans for under- skills requirements in an era of big data and cloud computing, taking a course, for example, they were 29% more likely to com- and constant employee churn that leaves the company having to plete the course than a control group who did not have to do so. fill 50,000 jobs a year. Recruiting from outside is difficult, expen- Information about effective learning strategies can be per- sive and liable to cause ill-feeling among existing staff. The firm’s sonalised, too. The Open University, a British distance-learning answer is an ambitious plan to reskill its own people. institution, already uses dashboards to monitor individual stu- Employees each have a career profile that they maintain dents’ online behaviour and performance. Knewton, whose themselves, which contains a record of their skills and training. platform captures data on 10m current American students, rec- They also have access to a database called “career intelligence”, ommends personalised content to them. Helping people to be which shows them the jobs on offer within the company, what more aware of their own thought processes when they learn skills they require and how much demand there is for them. The makes it more likely they can acquire new skills later in life. 7 firm hasdeveloped shortcoursescalled nanodegreeswith Udac- ity, the MOOC provider, and is also working with universities on developing course curriculums. Employees work in their own Upstarts and incumbents time to build theirskills. ButAT&T appliesboth carrotand stick to encourage them, bywayofgeneroushelp with tuition fees (total- ling $30m in 2015) for those who take courses and negative ap- The return of the MOOC praisal ratings forthose who show no interest. As continued learning becomes a corporate priority, two questions arise. First, is it possible for firms to screen candidates and employees on the basis of curiosity, or what psychologists call “need for cognition”? Getting through university is one very Alternative providers of education must solve the rough proxy for this sort of foundational skill, which helps ex-

plain why so many employers stipulate degrees for jobs which problems of cost and credentials

on the face ofit do not require them. T E O MOOCs peaked in 2012. Salman Khan, an investment analyst who had begun teaching bite-sized les- Curiouser and curiouser sons to his cousin in New Orleans over the internet and turned More data-driven approaches are also being tried. Man- that activity into a wildly popular educational resource called power, a human-resources consultancy, is currently running the Khan Academy, was splashed on the cover of Forbes. Sebas- trials on an app that will score individuals on their “learn- tian Thrun, the founder of another MOOC called Udacity, pre- ability”. Knack, a startup, offers a series of apps that are, in effect, dicted in an interview in Wired magazine that within 50 years gamified psychological tests. In Dashi Dash, for example, partici- the numberofuniversitieswould collapse to justten worldwide. pants play the part of waiters and are asked to take the orders of The New York Times declared it the year ofthe MOOC. customers on the basis of (often hard to read) expressions. As The sheer numbers ofpeople flocking to some ofthe initial more and more customers arrive, the job of managing the work- courses seemed to suggest that an entirely new model of open- flow gets tougher. Every decision and every minute change in access, free university education was within reach. Now MOOC strategy is captured as a data point and sent to the cloud, where sceptics are more numerous than believers. Although lots of machine-learning algorithms analyse players’ aptitudes against people still sign up, drop-out rates are sky-high. a reference population of 25,000 people. An ability to read ex- Nonetheless, the MOOCs are on to something. Education, pressions wins points for empathy; a decision always to serve like health care, is a complex and fragmented industry, which 1

The Economist January 14th 2017 5 SPECIAL REPORT LIFELONG EDUCATION

requirement for teachers to draw a pension from the industrial- ist’s nascent pension scheme forcollege faculty. Students in their early 20s can more easily afford a lengthy time commitment because they are less likely to have other responsibilities. Although millions of people do manage part-time or distance learning in later life—one-third of all working students currently enrolled in America are 30-54 years old, according to the Georgetown Uni- versity Centre on Education and the Workforce—balancing learning, working and family life can cause enormous pressures. Moreover, the world ofworkincreasingly demands a quick response from the education system to provide people with the desired qualifications. To take one example from Burning Glass, in 2014 just under 50,000 American job-vacancy ads asked for a CISSP cyber-security certificate. Since only 65,000 people in America hold such a certificate and it takes five years of experience to earn one, that requirement will be hard to meet. Less demanding professions also put up huge barriers to entry. If you want to become a licensed cosmetologist in New Hampshire, you will need to have racked up 1,500 hours oftraining. 2 makes it hard to gain scale. Despite those drop-out rates, the In response, the MOOCs have tried to make theircontent as MOOCs have shown it can be done quickly and comparatively digestible and flexible as possible. Degrees are broken into mod- cheaply. The Khan Academy has 14m-15m users who conduct at ules; modules into courses; courses into short segments. The least one learning activity with it each month; Coursera has 22m MOOCs test for optimal length to ensure people complete the registered learners. Those numbers are only going to grow. course; six minutes is thought to be the sweet spot for online vid- FutureLearn, a MOOC owned by Britain’s Open University, has eo and four weeks for a course. big plans. Oxford University announced in November that it Scott DeRue, the dean of the Ross School of Business at the would be producing its first MOOC on the edX platform. University ofMichigan, says the unbundling ofeducational con- In their search for a business model, some platforms are tent into smaller components reminds him of another industry: now focusing much more squarely on employment (though oth- music. Songs used to be bundled into albums before being disag- ers, like the Khan Academy, are not for profit). Udacity has gregated by iTunesand streaming services such as Spotify. In Mr launched a series of nanodegrees in tech-focused courses that DeRue’s analogy, the degree is the album, the course content that range from the basic to the cutting-edge. Ithasdone so, moreover, is freely available on MOOCs is the free streaming radio service, in partnership with employers. A course on Android was devel- and a “microcredential” like the nanodegree or the specialisa- oped with Google; a nanodegree in self-drivingcars uses instruc- tion is paid-foriTunes. tors from Mercedes-Benz, Nvidia and others. Students pay $199- How should universities respond to that kind of disrup- 299 a month for as long as it takes them to finish the course (typi- tion? For his answer, Mr DeRue again draws on the lessons ofthe cally six to nine months) and get a 50% rebate if they complete it music industry. Faced with the disruption caused by the internet, within a year. Udacity also offers a souped-up version of its na- it turned to live concerts, which provided a premium experience nodegree for an extra $100 a month, along with a money-back that cannot be replicated online. The on-campus degree also guarantee ifgraduates do not find a job within six months. needs to markitselfout as a premium experience, he says. Coursera’scontentcomeslargelyfrom universities, not spe- Another answer is for universities to make their own pro- cialist instructors; its range is much broader; and it is offering full ducts more accessible by doing more teaching online. This is be- degrees (one in computer science, the other an MBA) as well as ginning to happen. When Georgia Tech decided to offer an on- shorter courses. But it, too, has shifted its emphasis to employ- line version of its masters in computer science at low cost, many ability. Its boss, Rick Levin, a former president of Yale University, were shocked: it seemed to risk cannibalising its campus degree. cites research showing that half of its learners took courses in or- But according to Joshua Goodman of Harvard University, who der to advance their careers. Although its materials are available has studied the programme, the decision was proved right. The without charge, learners pay forassessment and accreditation at campus degree continued to recruit students in their early 20s the end of the course ($300-400 for a four-course sequence that whereasthe online degree attracted people with a median age of Coursera calls a “specialisation”). It has found that when money 34 who did not want to leave their jobs. Mr Goodman reckons is changing hands, completion rates rise from 10% to 60% . It is in- this one programme could boost the numbers of computer-sci- creasingly working with companies, too. Firms can now inte- ence masters produced in America each year by 7-8%. Chip Pau- grate Coursera into their own learning portals, track employees’ cek, the boss of2U, a firm that creates online degree programmes participation and provide their desired menu ofcourses. for conventional universities, reports that additional marketing These are still early days. Coursera does not give out figures efforts to lure online students also boost on-campus enrolments. on its paying learners; Udacity says it has13,000 people doing its nanodegrees. Whatever the arithmetic, the reinvented MOOCs Educational Lego matter because they are solving two problems they share with Universities can become more modular, too. EdX has a mi- every provider oflater-life education. cromastersin supply-chain managementthatcan eitherbe taken The first of these is the cost of learning, not just in money on its own or count towards a full masters at MIT. The University but also in time. Formal education rests on the idea of qualifica- of Wisconsin-Extension has set up a site called the University tions that take a set period to complete. In America the en- LearningStore, which offers slivers ofonline content on practical trenched notion of “seat time”, the amount of time that students subjects such as project management and business writing. En- spend with school teachers or university professors, dates back thusiasts talkofa world of“stackable credentials” in which qual- to Andrew Carnegie. It was originally intended as an eligibility ifications can be fitted together like bits ofLego. 1

6 The Economist January 14th 2017 SPECIAL REPORT LIFELONG EDUCATION

2 Just how far and fast universities will go in this direction is something faintly regressive about the world of microcreden- unclear, however. Degrees are still highly regarded, and in- tials. Like a university degree, it still involves a stamp ofapproval creased emphasison critical thinkingand social skills raisestheir from a recognised provider after a proprietary process. Yet lots of value in many ways. “The model of campuses, tenured faculty learning happens in informal and experiential settings, and lots and so on does not work that well for short courses,” adds Jake ofworkplace skills cannot be acquired in a course. Schwartz, General Assembly’s boss. “The economics ofcovering fixed costs forces them to go longer.” Gold stars for good behaviour Academic institutions also struggle to deliver really fast- One way of dealing with that is to divide the currency of moving content. Pluralsight uses a model similar to that of book knowledge into smaller denominations by issuing “digital publishing by employing a network of1,000 experts to produce badges” to recognise less formal achievements. RMIT University, and refresh itslibraryofvideoson ITand creative skills. These ex- Australia’s largest tertiary-education institution, is working with perts get royalties based on how often their content is viewed; its Credly, a credentialling platform, to issue badges for the skills highest earner pulled in $2m last year, according to Aaron Skon- that are not tested in exams but that firms nevertheless value. Be- nard, the firm’s boss. Such rewards provide an incentive for au- linda Tynan, RMIT’s vice-president, cites a project carried out by thors to keep updatingtheircontent. University faculty have oth- engineering students to build an electric car, enter it into races er priorities. and win sponsors as an example. Beside costs, the second problem for The trouble with digital badges is MOOCs to solve is credentials. Close col- that they tend to proliferate. Illinois State leagues know each other’s abilities, but University alone created 110 badges when modern labour markets do not work on it launched a programme with Credly in the basis of such relationships. They need 2016. Add in MOOC certificates, LinkedIn widely understood signals of experience Learningcourses, competency-based edu- and expertise, like a university degree or a cation, General Assembly and the like, baccalaureate, however imperfect they and the idea ofcreating new currencies of may be. In their own fields, vocational knowledge starts to look more like a reci- qualifications do the same job. The pe for hyperinflation. MOOCs’ answer is to offer microcreden- David Blake, the founderofDegreed, tials like nanodegrees and specialisations. a startup, aspires to resolve that problem But employers still need to be confi- by acting as the central bank of creden- dent that the skills these credentials tials. He wants to issue a standardised vouchsafe are for real. LinkedIn’s “en- assessment of skill levels, irrespective of dorsements” feature, for example, was how people got there. The plan is to create routinely used by members to hand out a network of subject-matter experts to as- compliments to people they did not know sess employees’ skills (copy-editing, say, for skills they did not possess, in the hope or credit analysis), and a standardised of a reciprocal recommendation. In 2016 grading language that means the same thing to everyone, everywhere. Pluralsight is heading in a similar di- People are more likely to invest in training if it confers a rection in its field. A diagnostic tool uses a technique called item response theory to qualification that others will recognise. But they also workout users’ skill levels in areas such as need to know which skills are useful in the first place coding, giving them a rating. The system helps determine what individuals should learn next, but also gives companies a the firm tightened things up, but getting the balance right is hard. standardised way to evaluate people’s skills. Credentials require just the right amount of friction: enough to A system of standardised skills measures has its own pro- be trusted, not so much as to blockcareer transitions. blems, however. Using experts to grade ability raises recursive Universities have no trouble winning trust: many of them questions about the credentials of those experts. And it is hard can call on centuriesofexperience and name recognition. Cours- for item response theory to assess subjective skills, such as an era relies on universities and business schools formost ofits con- abilityto constructan argument. Specific, measurable skills in ar- tent; their names sit proudly on the certificates that the firm is- eas such as IT are more amenable to this approach. sues. Some employers, too, may have enough kudos to play a So amenable, indeed, that they can be tested directly. As an role in authenticating credentials. The involvement of Google in adolescent in Armenia, Tigran Sloyan used to compete in mathe- the Android nanodegree has helped persuade Flipkart, an Indi- matical Olympiads. That experience helped him win a place at an e-commerce platform, to hire Udacity graduates sight unseen. MIT and also inspired him to found a startup called CodeFights Wherever the content comes from, students’ work usually in San Francisco. The site offers free gamified challenges to needs to be validated properly for a credential to be trusted. 500,000 users as a way of helping programmers learn. When When student numbers are limited, the marking can be done by they know enough, they are funnelled towards employers, the teacher. But in the world of MOOCs those numbers can spi- which pay the firm 15% ofa successful candidate’s starting salary. ral, making it impractical for the instructors to do all the assess- Sqore, a startup in Stockholm, also uses competitions to screen ments. Automation can help, but does not work for complex as- job applicants on behalfofits clients. signments and subjects. Udacity gets its students to submit their However it is done, the credentialling problem has to be coding projects via GitHub, a hosting site, to a network of mach- solved. People are much more likely to invest in training ifit con- ine-learning graduates who give feedbackwithin hours. fers a qualification that others will recognise. But they also need Even if these problems can be overcome, however, there is to know which skills are useful in the first place. 7

The Economist January 14th 2017 7 SPECIAL REPORT LIFELONG EDUCATION

Career planning positions in industries like retailing and health care. The programme starts by going into workplaces and identifying key events (how an IT helpdesk handles a call from an irate custom- Pathway dependency er, for example) that distinguish high performers from the rest. Curriculum designers then use that analysis to create a full- time training programme lasting between four and 12 weeks that covers both technical knowledge and behavioural skills. The programme has gone live in America, Spain, India, Kenya and How to turn a qualification into a salary Mexico. By the end of 2016 it had 10,000 graduates, for whom it claims an employment rate of 90% and much higher retention UNIVERSITY EDUCATION IS designed to act as a slipway, rates than usual. The trainees pay nothing; the hope is that em- launching students into the wider world in the expectation ployers will fund the programme, or embed it in their own train- that the currents will guide them into a job. In practice, many ing programmes, when they see how useful it is. people getstuckin the doldrumsbecause employersdemand ev- idence of specific experience even from entry-level candidates. A little help from your friends Whether this counts as a skills gap is a matter of debate. “If I can- Such experiments use training to take people into specific not find a powerful, fuel-efficient, easy-to-park car for $15,000, jobs. In the past, an initial shove might have been all the help that doesn’t mean there is a car shortage,” says Peter Cappelli of they needed. But as middle-skilled roles disappear, some rungs the Wharton School of the University of Pennsylvania. But on the job ladder have gone missing. And in a world of continu- whether the fault lies with the educators or the employers, there ous reskilling and greater self-employment, people may need is a need for pathways that lead individuals into jobs. help with repeatedly moving from one type of job to another. Sometimes those pathways are clearly defined, as in medi- Vocational education is good at getting school-leavers into work, cine and the law. Vocational education combines classroom and but does nothing to help people adapt to changes in the world of work-based learning to prepare young people for specific trades. work. Indeed, a cross-country study in 2015 by researchers at the In many European countries, one-third to half of later-stage sec- HooverInstitution suggests that people with a vocational educa- ondary schoolgoers are on a vocational path (see chart). Britain tion are more likely than those with a general education to with- is due to introduce an apprenticeship levy in April. drawfrom the labourforce astheyage. The pattern isparticularly But pathways are needed to smooth transitions in other marked in countriesthatrelyheavilyon apprenticeships, such as countries (America, for example, lacks a tradition of vocational Denmark, Germany and Switzerland. education); in less structured occupations; and when formal Large companies may have the scale to offer their employ- education has come to an end. The nanodegree is an example of ees internal pathways to improve their skills, as companies like such a pathway, asisGeneral Assembly’sbootcamp model. Both AT&T do. But many workers will need outside help in deciding rely heavily on input from employers to create content; both use which routes to take. That suggests a big opportunity for firms jobs rather than credentials as a measure ofsuccess. that can act, in effect, as careers advisers. Some are better placed That is particularly important in the early stages of people’s than others to see where the jobs market is going. Manpower, careers, which is not just when they lack experience but also which supplies temporary workers to many industries, last year when earnings grow fastest. An analysis of American wage launched a programme called MyPath that is based on the idea growth by economists at the New York Federal Reserve showed of an iterative process of learning and working. It allows Man- that the bulk of earnings growth took place between the ages of power’s army oftemporary workers in America to earn a degree 25 and 35; on average, after the age of 45 only the top 2% of life- from Western International University at no financial cost to time earners see any earnings growth. So it is vital for people to them. The degree is structured as a series of three or four epi- move quickly into work once qualified, and to hold on to jobs sodes of education followed by periods in work, in the expecta- once they get them. tion that Manpower has a good overview of the skills leading to That is the insight behind well-paid jobs. LearnUp, a startup that works LinkedIn is another organisation with a decent under- with applicants without col- standing of wider trends. The professional-networking site likes lege degrees for entry-level po- to call the data it sits on “the economic graph”, a digital map of sitions. Users applying for a job the global economy. Its candidate data, and its recruitment plat- online can click on a link and form, give it information on where demand from employers is take a one-hour online training greatest and what skills jobseekers need. And with LinkedIn session on how to be a cashier, Learning it can now also deliver training itself. sales clerkor whatever they are The firm can already tell candidates how well their qualifi- after. Employers pay LearnUp a cations forany advertised job stack up against those of other ap- fixed fee to improve the pool of plicants. In time, its data might be used to give “investment ad- candidates. Recruitment and vice”, counselling its members on the financial return to specific retention rates have risen. skills and on how long they are likely to be useful; or to show Generation, a philan- members how other people have got into desirable positions. thropically funded programme The difficulty with offering mass-market careers advice is run by the McKinsey Social Ini- finding a business model that will pay forit. LinkedIn solves this tiative, a not-for-profit arm of problem by aiming itself primarily at professionals who either the consultancy, uses a boot- pay for services themselves or who are of interest to recruiters. camp approach and some typi- But that raises a much bigger question. “There is no shortage of cally McKinsey-esque thinking options for folks of means,” says Adam Newman of Tyton Part- to train people from difficult ners, an education consultancy. “But what about LinkedIn for backgrounds formiddle-skilled the linked-out?” 7

8 The Economist January 14th 2017 SPECIAL REPORT LIFELONG EDUCATION

tion of sensors on motorways, that might prompt retraining. “Thisisa boiling-frogproblem,” he says. “Itisnotthoughtabout.” For lower-skilled workers of this sort the world of MOOCs, General Assembly and LinkedIn is a million miles away. Around 80% of Coursera’s learners have university degrees. The costs of reskilling, in terms oftime and money, are easiest to bear forpeo- ple who have savings, can control their working hours or work for companies that are committed to upgrading their workforce. And motivation is an issue: the tremendous learning opportunities offered by the internet simply do not appeal to everyone. Whosoever hath not The rewards of retraining are highest for computing skills, but there is no natural pathway from trucker to coder. And even if there were, many of those already in the workforce lack both the confidence and the capability to make the switch. In its Pro- gramme for the International Assessment of Adult Competen- cies, the OECD presents a bleakpicture ofskills levels in 33 mem- ber countries (see chart). One in five adults, on average, has poor reading and numeracy skills. One in four has little or no experience of computers. On a measure of problem-solving ability using technology, most adults are at or below the lowest level of proficiency. Low-skilled workers Moreover, learning is most effective when people are able to practise their new skills. Yet many jobs, including lorry-driving, afford little such opportunity, and some of them are being The elephant in the deskilled further. Research by Tom Higgins of Cardiff University suggests that the numeracy requirements forretail assistants and truck care-home workersin Britain wentdown between 1997 and 2012. The head ofone ofthe world’s biggest banks worries that a back- office operation in India has disaggregated its workinto separate The emerging system of lifelong learning will do little tasks so effectively that employees are no longer able to under- stand the processesasa whole, letalone make useful suggestions to reduce inequality for improving them. IMAGINE YOU ARE a 45-year-old long-distance lorry So the truckers’ dilemma will be very hard to solve. “It’s dif- driver. You never enjoyed school and left as soon as you ficult when you don’t have a good answer even in an ideal could, with a smattering of qualifications and no great love of world,” says Jesper Roine, an economist who sat on a Swedish learning. The job is tiring and solitary, but it does at least seem to commission to examine the future of work. But as a thought ex- offer decent job security: driver shortages are a perennial com- periment it highlights some of the problems involved in upgrad- plaint in the industry, and the average age of the workforce is ing the stock of low-skilled and mid-skilled workers. Any decent high (48 in Britain), so the shortfalls are likely to get worse. Amer- answer will need a co-ordinated effort to bring together individ- ica’s Bureau of Labour Statistics (BLS) says there were 1.8m truck- uals, employers and providers of education. That suggests a role ers in 2014 and expects a 5% rise in their number by 2024. “As the for two entities in particular. economy grows, the demand for goods will increase and more One is trade unions. They have an industry-wide view of truckdrivers will be needed to keep supply chains moving,” pre- trends that may not be available to smaller employers. They can dicts the BLS website, chirpily. also accompany people throughout their working lives, which 1 But the future might unfold very differently. For all the ex- citement over self-driving passenger cars, the freight industry is likely to adopt autonomous vehicles even faster. And according to a report in 2014 by Morgan Stanley, a bank, full automation might reduce the pool of American truck drivers by two-thirds. Those projections came hedged with caveats, and rightly so. The pace of adoption may be slowed by regulation. Drivers may still be needed to deal with unforeseen problems; ifsuch jobs require more technical knowledge, they may even pay better. Employ- ment in other sectors may grow as freight costs come down. But there is a chance that in the not too distant future a very large numberoftruckerswill find themselvesredundant. The implica- tions are immense. Knowing when to jump is one problem. For people with decades of working life still ahead of them, it is too early to quit but it is also risky to assume that nothing will change. Matthew Robb of Parthenon-EY, a consultancy, thinks that governments should be talkingto industry bodies about the potential formass redundancies and identifying trigger points, such as the installa-

The Economist January 14th 2017 9 SPECIAL REPORT LIFELONG EDUCATION

2 may become increasingly important in a world of rising self-em- ones, to club together to signal ployment. Denmark’s tripartite system, for example, binds to- their skills needs to the work- getheremployers, governmentand unions. Firmsand unions get force at large. Individual learn- together to identify skills needs; collective-bargaining agree- ing accounts have a somewhat ments enshrine rights to paid leave for training. The country’s chequered history—fraudulent famed “flexicurity” system offers unemployed workers a list of training providers helped scup- 258 vocational-training programmes. per a British experiment in the In Britain a well-regarded programme called UnionLearn early 2000s—but if well de- uses union representatives both to inform workers about train- signed, they can offer workers ingoptions and to liaise with employers on workers’ requests for educational opportunities with- training. Employees seem more likely to discuss shortfalls in ba- out being overly prescriptive. sic skills with union representatives than with managers. An analysis by academics at Leeds University Business School Any fool can know shows that between 2001 and 2013 union members in Britain In June 2016, this newspa- were a third more likely to have received training than non- per surveyed the realm of artifi- unionised workers. cial intelligence and the adjust- The second entity is government. There is much talk about ments it would require workers lifelong learning, though few countries are doing much about it. to make as jobs changed. “That The Nordics fall into this less populated camp. But it is Singapore will mean making education that can lay claim to the most joined-up approach with its Skills- and training flexible enough to Future initiative. Employers in the city-state are asked to spell out teach new skills quickly and effi- the changes, industry by industry, that they expect to happen ciently,” we concluded. “It will over the next three to five years, and to identify the skills they require a greater emphasis on will need. Theiranswersare used to create “industrytransforma- lifelong learning and on-the-job Mass entertainment February 11th tion maps” designed to guide individuals on where to head. training, and wider use of on- The future of Europe March 25th The Pearl River delta April 8th Since January 2016 every Singaporean above the age of 25 line learning and video-game- hasbeen given a S$500 ($345) creditthatcan be freelyused to pay style simulation.” for any training courses provided by 500 approved providers, in- The uncertainties around cluding universities and MOOCs. Generous subsidies, of up to the pace and extent of techno- 90% for Singaporeans aged 40 and over, are available on top of logical change are enormous. this credit. The programme currently has a budget of S$600m a Some fear a future of mass un- year, which isdue to rise to S$1billion within three years. Accord- employment. Others are san- ing to Ng Cher Pong, SkillsFuture’s chiefexecutive, the returns on guine that people will have time to adapt. Companies have to thatspendingmatterlessthan changingthe mindsetaround con- wantto adoptnewtechnologies, afterall, and regulators mayim- tinuous reskilling. pede their take-up. What is not in doubt is the need for new and Some programmes cater to the needs of those who lack ba- more efficient ways to develop and add workplace skills. sic skills. Tripartite agreements between unions, employers and The faint outlines of a new ecosystem for connecting em- government lay out career and skills ladders for those who are ployment and education are becoming discernible. Employers trapped in low-wage occupations. Professional-conversion pro- are puttinggreateremphasis on adaptability, curiosity and learn- grammesoffersubsidised trainingto people switchingto newca- ing as desirable attributes foremployees. They are working with reers in areas such as health care. universities and alternative providers to create and improve Given Singapore’s size and political system, this approach their own supply of talent. Shorter courses, lower costs and on- is not easily replicated in many other countries, but lessons can line deliveryare makingiteasierforpeople to combine workand still be drawn. It makes sense for employers, particularly smaller training. New credentials are being created to signal skills.

At the same time, new technologies should make learning

more effective as well as more necessaryirtual and augmented The outlines reality could radically improve professional training. Big data offer the chance for more personalised education. Platforms make of a new it easier to connect people of differing levels of knowledge, al- ecosystem lowing peer-to-peer teaching and mentoring. “Education is be- comingflexible, modular, accessible and affordable,” says Simon for Nelson, the boss ofFutureLearn, the Open University MOOC. connecting But for now this nascent ecosystem is disproportionately likely to benefit those who least need help. It concentrates on ad- employment vanced technological skills, which offer the clearest returns and and are relatively easy to measure. And it assumes that people have the money, time, motivation and basic skills to retrain. education Thanks to examples like Singapore’s, it is possible to imag- are ine ways in which continuous education can be made more accessible and affordable for the mass of citizens. But it is as easy— becoming indeed, easier—to imagine a future in which the emerging infra- discernible structure of lifelong learning reinforces existing advantages. Far from alleviating the impact of technological upheaval, that would riskexacerbating inequality and the social and economic tensions it brings in its wake. 7

10 The Economist January 14th 2017